Incremental Majorization-Minimization Optimization with Application to Large-Scale Machine Learning

Julien Mairal 1, *
* Auteur correspondant
1 LEAR - Learning and recognition in vision
Inria Grenoble - Rhône-Alpes, LJK - Laboratoire Jean Kuntzmann, INPG - Institut National Polytechnique de Grenoble
Abstract : Majorization-minimization algorithms consist of successively minimizing a sequence of upper bounds of the objective function. These upper bounds are tight at the current estimate, and each iteration monotonically drives the objective function downhill. Such a simple principle is widely applicable and has been very popular in various scientific fields, especially in signal processing and statistics. In this paper, we propose an incremental majorization-minimization scheme for minimizing a large sum of continuous functions, a problem of utmost importance in machine learning. We present convergence guarantees for non-convex and convex optimization when the upper bounds approximate the objective up to a smooth error; we call such upper bounds ''first-order surrogate functions''. More precisely, we study asymptotic stationary point guarantees for non-convex problems, and for convex ones, we provide convergence rates for the expected objective function value. We apply our scheme to composite optimization and obtain a new incremental proximal gradient algorithm with linear convergence rate for strongly convex functions. In our experiments, we show that our method is competitive with the state of the art for solving machine learning problems such as logistic regression when the number of training samples is large enough, and we demonstrate its usefulness for sparse estimation with non-convex penalties.
Liste complète des métadonnées


https://hal.inria.fr/hal-00948338
Contributeur : Julien Mairal <>
Soumis le : vendredi 24 avril 2015 - 11:51:34
Dernière modification le : mercredi 14 décembre 2016 - 01:07:12
Document(s) archivé(s) le : lundi 14 septembre 2015 - 13:12:09

Fichier

95763.pdf
Fichiers éditeurs autorisés sur une archive ouverte

Licence


Copyright (Tous droits réservés)

Identifiants

Citation

Julien Mairal. Incremental Majorization-Minimization Optimization with Application to Large-Scale Machine Learning. SIAM Journal on Optimization, Society for Industrial and Applied Mathematics, 2015, 25 (2), pp.829-855. <http://epubs.siam.org/toc/sjope8/25/2>. <10.1137/140957639>. <hal-00948338v5>

Partager

Métriques

Consultations de
la notice

304

Téléchargements du document

145